Principal component analysis and neural network combining PCA provides an adaptive neural network parallel line of the main element and master spatial analysis techniques, but the data for non-Gaussian stochastic systems based on the minimum mean square error of the reconstructed PCA extracted . the main element direction than the direction of maximizing the information paper first presents a minimum mean square error based on self-association of the main element network analysis of variance reconstruction of its best properties of both IT and non-maximizing features; then presented with minimal residual Poor information entropy learning objectives PCA neural network and gives the approximate calculation method residual entropy network output and network learning methods; finally analyzed in Gaussian random distribution system, the minimum residual entropy and minimum mean square error of reconstruction The results are consistent.
|Translated title of the contribution||Comparisons between minimum error entropy and minimum mean squared error based on PCA neural networks|
|Number of pages||7|
|Journal||Pattern Recognition and Artificial Intelligence|
|Publication status||Published - Feb 2005|
- PCA neural networks
- minimum residual entropy
- minimum mean squared error